|
Post by Soccerhouse on Aug 26, 2015 13:26:34 GMT -5
On the relative age effect and a new U.S. Soccer youth rule change www.topdrawersoccer.com/the91stminute/2015/08/on-the-relative-age-effect-and-a-new-u-s-soccer-youth-rule-change/"But the wider question is how, exactly, this is going to change anything in regard to relative age effect." Another read also from Oct 2014 offballmovement.blogspot.com/2014/10/exploring-age-bias-in-us-based-mls.htmlSimiliar to what zizou did for ecnl, article did the same on US-based MLS academies - "It is pretty obvious that the answer is yes, MLS academies have a bias toward selecting older players. January and February are the birth months with the most academy players. November and December had the least. In fact, Those born in the first two months of the year were 2.4 times more likely to find a roster spot than those born in the last two months. However, there’s another story told in this histogram. There’s a significant jump between July and August. In the part of youth soccer that isn’t USSDA, the age cutoffs aren’t simply based on calendar years. They are usually right there at the July/August split causing the older players in an age group to be born in August and the youngest in July. This spike in August could tell us a couple different stories. There are two that I think are more likely than others. The first is that the older players (August born) stand out more due to their physical traits in games where MLS academy coaches scout to recruit players. The second is that the older players are more likely to make rosters of non-USSDA “elite” club teams and get more playing time there resulting in them being more likely to develop soccer skills. It is impossible to know which is the larger contributor, but I suspect both are true to some extent."
|
|
|
Post by youthsoccerdad on Aug 26, 2015 13:34:13 GMT -5
Well the way it works now is that the Jan - Mar kids are able to make the top team and get to compete against older kids so when they ultimately play against their birth year in DA it seemed relatively "easy" and thus more make the team. The Aug-Oct kids get pampered in Academy with "better" positions such as playing center midfield for the entire game and have their confidence built playing against younger kids.
Everyone knows that this happens at a macro level. It is not easily accounted for now and it will not be later. It is what it is.
Here is where the real age bias sticks out - it is not that more kids are picked in January and February. It is that the kid born in June/July. What happens? They get put on the second/third/forth team, play against weaker competition. Get stuck being defender more. Coaches yell at them more often, they lose confidence.
Anyways, that is the part of this that always suprises me is that folks talk about the kids selected but look at June/July and you see the real bias.
|
|
|
Post by jash on Aug 26, 2015 13:54:32 GMT -5
If I'm reading this right, the current school-year based cutoff is helping to smooth the relative age effect. That's very interesting, if true, and is a direct piece of evidence saying the change will have a negative effect overall.
|
|
|
Post by Soccerhouse on Aug 26, 2015 14:03:21 GMT -5
that's how I read it also, now you will lose that effect and continue the linear trend down.
|
|
|
Post by youthsoccerdad on Aug 26, 2015 14:07:54 GMT -5
If I'm reading this right, the current school-year based cutoff is helping to smooth the relative age effect. That's very interesting, if true, and is a direct piece of evidence saying the change will have a negative effect overall. It just changes the distribution of players and the curve. The new system will provide greater value to the kid born in June or July than the old one, their bar would probably go up closer to what May looks like.
|
|
|
Post by zizou on Aug 26, 2015 14:15:30 GMT -5
I have something even more shocking to show in the ECNL data but I am scared people will totally come undone so I have been keeping it to myself. How is that for a tease?
|
|
|
Post by Soccerhouse on Aug 26, 2015 14:21:37 GMT -5
I have something even more shocking to show in the ECNL data but I am scared people will totally come undone so I have been keeping it to myself. How is that for a tease? HA love it! Will have to create a new tease award for this one!!
|
|
|
Post by soccerdad44 on Aug 26, 2015 14:50:29 GMT -5
that's how I read it also, now you will lose that effect and continue the linear trend down. Yes, I think that's right. The later you're born in the year, the less likely to play academy, and if you do, the less likely to be on a top team.
|
|
|
Post by jash on Aug 26, 2015 15:01:39 GMT -5
If I'm reading this right, the current school-year based cutoff is helping to smooth the relative age effect. That's very interesting, if true, and is a direct piece of evidence saying the change will have a negative effect overall. It just changes the distribution of players and the curve. The new system will provide greater value to the kid born in June or July than the old one, their bar would probably go up closer to what May looks like. That isn't the way I read it. There is a natural selection bias right now toward January-March players because that is what the highest level uses for cutoffs. But on those teams there is a bump for August-September players that would not otherwise be there if those players weren't getting the natural selection bias in club soccer, and the advantages that come with it. When the two systems align their dates, the outcome will look more like this (which is how it is in Europe, with everything aligned):
|
|
|
Post by rifle on Aug 26, 2015 15:34:10 GMT -5
I have something even more shocking to show in the ECNL data but I am scared people will totally come undone so I have been keeping it to myself. How is that for a tease? do tell...
|
|
|
Post by zizou on Aug 26, 2015 16:08:04 GMT -5
I have something even more shocking to show in the ECNL data but I am scared people will totally come undone so I have been keeping it to myself. How is that for a tease? do tell... These are the ECNL data for last year's players by birthday month. You see two age groups listed, U14 and U18. "Playing Up" means a player was something younger than U14 or U18, respectively. Many of those U18 play-ups are accounted for by a few clubs that seem to use younger aged players liberally, but that does not account for the totality of the dramatic increase at U18. There are about 1500 players in each age group. The red line is the US Census birth rates from the mid 2000s, and provides some baseline against which to compare the bars. I will leave this for open discussion, but, at least in this one elite league, things do not appear to work out well for the players with birth months early in the annual cycle. And even though there is some "recovery" for players born late in the annual cycle, they never fully reach the level of participation based on US Census data.
|
|
|
Post by jash on Aug 26, 2015 17:35:45 GMT -5
These are the ECNL data for last year's players by birthday month. You see two age groups listed, U14 and U18. "Playing Up" means a player was something younger than U14 or U18, respectively. Many of those U18 play-ups are accounted for by a few clubs that seem to use younger aged players liberally, but that does not account for the totality of the dramatic increase at U18. There are about 1500 players in each age group. The red line is the US Census birth rates from the mid 2000s, and provides some baseline against which to compare the bars. I will leave this for open discussion, but, at least in this one elite league, things do not appear to work out well for the players with birth months early in the annual cycle. And even though there is some "recovery" for players born late in the annual cycle, they never fully reach the level of participation based on US Census data. Not really sure what to do with this one. At U14 it looks pretty much like you'd expect. I don't know much about ECNL or the players' habits. What about intervening age groups? Could the huge drop-off of players be the 'best' players quitting after committing to college so they don't get injured?
|
|
|
Post by sidelinemama on Aug 26, 2015 17:57:57 GMT -5
I think it illustrates the fact that size evens out as girls get a bit older. I think there could be some burn out factor too from girls that peak at an early age that worked really hard to get there.
|
|
|
Post by zizou on Aug 28, 2015 10:55:15 GMT -5
Could the huge drop-off of players be the 'best' players quitting after committing to college so they don't get injured? That is a theory, but one that I think is not consistent with the available data. All you need to know about ECNL for present purposes is that 90% of the graduating players go on to play college soccer and that 90% of those players play D1. I have added the U16s to this version of the plot. Data from about 1500 players in each group. The oldest players are losing their advantage by this time and the youngest players are starting to gain. There are other theories that probably can handle such data more efficiently. I also have a sinister theory for the October birthday bump at U16 but I will keep that to myself for the present!
|
|
|
Post by paterfamilias on Aug 28, 2015 11:18:10 GMT -5
There are lies, damn lies and then there are statics... quoted by many
With that said tho, that is fantastic data full of great statistics.
I would be interested to see the burnout rate of college players after their first year. I think that would make for an interesting next step discussion
|
|
|
Post by Soccerhouse on Aug 28, 2015 11:19:27 GMT -5
Its hard to tell why, but it seems we are talking about 1 to 2 players movement etc per team. 1 or 2 less Aug/Sept birthdays shifted to maybe 1 may birthday. I could see that on every team, you have 1 aug/sept kid that by the time they hit u18 are no longer the player they were at u14 etc or by the time they reach this age 1 or 2 spring birthdays have reached full maturity and are now rostered. On a 18 man roster, each person is about .055, which is nuts because this graph highlights the dramatic effect 1 or 2 players from each team can make on the entire ECNL distribution.
|
|
|
Post by jash on Aug 28, 2015 12:46:39 GMT -5
Could the huge drop-off of players be the 'best' players quitting after committing to college so they don't get injured? That is a theory, but one that I think is not consistent with the available data. All you need to know about ECNL for present purposes is that 90% of the graduating players go on to play college soccer and that 90% of those players play D1. I have added the U16s to this version of the plot. Data from about 1500 players in each group. The oldest players are losing their advantage by this time and the youngest players are starting to gain. There are other theories that probably can handle such data more efficiently. I also have a sinister theory for the October birthday bump at U16 but I will keep that to myself for the present! What is most unusual about this is that in other groups where age selection bias exists, there is no such shift through the years. Data I've looked at for European leagues shows the age selection bias continues unhindered going forward. Pretty much all of that data is for males though. There is some external factor at work here, given how incredibly strong the age selection bias is at U14. It might have to do with gender (though that seems unlikely), or it might have to do with college, or it could be something else. But there has to be a reason for a reversal of the selection bias, and such anomalies usually come down to some unaccounted-for external factor.
|
|
|
Post by zizou on Aug 28, 2015 13:06:14 GMT -5
What is most unusual about this is that in other groups where age selection bias exists, there is no such shift through the years. Data I've looked at for European leagues shows the age selection bias continues unhindered going forward. Pretty much all of that data is for males though. There is some external factor at work here, given how incredibly strong the age selection bias is at U14. It might have to do with gender (though that seems unlikely), or it might have to do with college, or it could be something else. But there has to be a reason for a reversal of the selection bias, and such anomalies usually come down to some unaccounted-for external factor. Let's see an example of the European data your talking about plotted by playing year.
|
|
|
Post by jash on Aug 28, 2015 13:17:02 GMT -5
What is most unusual about this is that in other groups where age selection bias exists, there is no such shift through the years. Data I've looked at for European leagues shows the age selection bias continues unhindered going forward. Pretty much all of that data is for males though. There is some external factor at work here, given how incredibly strong the age selection bias is at U14. It might have to do with gender (though that seems unlikely), or it might have to do with college, or it could be something else. But there has to be a reason for a reversal of the selection bias, and such anomalies usually come down to some unaccounted-for external factor. Let's see an example of the European data your talking about plotted by playing year. astro1.panet.utoledo.edu/~ljc/birth_sports.pdf
|
|
|
Post by Soccerhouse on Aug 28, 2015 13:21:06 GMT -5
The third solution is great!
A third solution would be to change the mentality of youth team coaches (Barnsley & Thompson, 1988; Helsen et al., 2000a,b). Coaches should pay more attention to technical and tactical skills when selecting players as opposed to an over-reliance on physical characteristics such as height. In a similar vein, coaches should be encouraged to change their philosophical approach to instruction. The statement that ‘‘winning isn’t everything, but the only thing’’ currently represents the strategic thinking of many youth coaches. Coaches should find a better balance between short-term success and a more task- or process-oriented approach to instruction. Clearly, ‘‘winning’’ does matter at the elite level in soccer. In this sense, the players must be exposed to such a reality at some stage during their progression to the elite level. This is perhaps especially important when selecting national youth teams. In the professional club teams, player development should hopefully be viewed as a more long-term process spanning a 10- year period and beyond. In any case, it would be big step forward if the philosophy of future coaches in general, and of those who are involved in the professional clubs in particular, may be more guided by the premise that ‘‘there is more to coaching than just winning’’.
|
|
|
Post by zizou on Aug 28, 2015 13:39:02 GMT -5
Thanks Jash! So just a few quick observations. First, this study is using only NT players. Kind of a different sample. The ECNL players, in general, might be considered to be "Elite" (at least they say so in their name!), but only a handful of them are NT caliber. Big difference. Second, The effects were not statistically significant for the U18 Woman and the the U21 men. I did not go over this paper in excruciating detail (can do that later), but I am missing the dramatic contradiction between what is in this paper and what is in the ECNL data. One possibility is that the males could have a different time course for dissolution of the effect. It could be a bias in the data in this paper due to selection of such an extreme part of the distribution. That is a pretty rarified sample, my man.
|
|
|
Post by jash on Aug 28, 2015 13:51:20 GMT -5
Thanks Jash! So just a few quick observations. First, this study is using only NT players. Kind of a different sample. The ECNL players, in general, might be considered to be "Elite" (at least they say so in their name!), but only a handful of them are NT caliber. Big difference. Second, The effects were not statistically significant for the U18 Woman and the the U21 men. I did not go over this paper in excruciating detail (can do that later), but I am missing the dramatic contradiction between what is in this paper and what is in the ECNL data. One possibility is that the males could have a different time course for dissolution of the effect. It could be a bias in the data in this paper due to selection of such an extreme part of the distribution. That is a pretty rarified sample, my man. Oh sure, this was just the first one I put my hands on. If you want more data, there are probably 30 relevant articles linked from here: www.ncbi.nlm.nih.gov/pmc/articles/PMC3761747/
|
|
|
Post by Soccerhouse on Aug 28, 2015 13:57:59 GMT -5
Thanks Jash! So just a few quick observations. First, this study is using only NT players. Kind of a different sample. The ECNL players, in general, might be considered to be "Elite" (at least they say so in their name!), but only a handful of them are NT caliber. Big difference. Second, The effects were not statistically significant for the U18 Woman and the the U21 men. I did not go over this paper in excruciating detail (can do that later), but I am missing the dramatic contradiction between what is in this paper and what is in the ECNL data. One possibility is that the males could have a different time course for dissolution of the effect. It could be a bias in the data in this paper due to selection of such an extreme part of the distribution. That is a pretty rarified sample, my man. Oh sure, this was just the first one I put my hands on. If you want more data, there are probably 30 relevant articles linked from here: www.ncbi.nlm.nih.gov/pmc/articles/PMC3761747/very interesting! Makes me think even more that having 2 systems in place like we had one that used calendar year & one that uses academic is was a good solution......
|
|
|
Post by jash on Aug 28, 2015 14:29:32 GMT -5
very interesting! Makes me think even more that having 2 systems in place like we had one that used calendar year & one that uses academic is was a good solution...... Yep, that August bump really helped engage MORE players to be eligible to play. That has to be a good thing.
|
|
|
Post by soccerdad44 on Aug 28, 2015 15:42:32 GMT -5
Wow, I didn't know that being the youngest had an official name (Relative Age Effect). Now I see the benefit of having 2 systems in place. It likely boosted the number or people playing higher level soccer.
|
|
|
Post by zizou on Sept 4, 2015 12:28:49 GMT -5
Not really sure what to do with this one. At U14 it looks pretty much like you'd expect. I don't know much about ECNL or the players' habits. What about intervening age groups? Could the huge drop-off of players be the 'best' players quitting after committing to college so they don't get injured? I finished adding all the ECNL age groups. Instead of plotting proportions I plotted total number of births by month since there are about 1500 observations in each age group and Soccerhouse was interested in the raw number of players by month. There are around 8000 observations in this plot. There were 78 total clubs in the league last year, but not every team from every club provided birth date data. It is hard to argue there is not an age-graded drop-off in early birth month advantage in this one league of elite female soccer players, arguably the pre-eminent such league in the USA. There could be multiple reasons for this decline. Going to college cannot explain the decline given that it starts at U15/U16. Although it could account for some of the U18 effect if those players had already graduated high school. The thing that is really unfortunate about these data to me is that the youngest birth month players never really recover their numbers. They do a bit, but hardly enough to make up for the initial bias against their selection. What a terrible disservice has been perpetrated on players unlucky by virtue of their birth month. There are multiple differences between this population of players and the ones in the studies jash so kindly linked. But one thing that always occurs to me when I see findings that do not follow prevailing wisdom (happens all the time in my field) is that things are not so simple as the might initially appear.
|
|